1. Source selection and scope
Inputs include public social signals, official sites, product blogs, public webpages, newsletters, and APIs. Entering collection scope does not make a source item publish automatically.
How we collect, select, and revise AI intelligence
Automation processes public signals continuously. Editors remain responsible for selection, verification, significance, and the final publication.
This page documents the current workflow and its boundaries. Summaries and translations improve reading efficiency but do not replace original sources or claim to archive an entire platform or the public web.
- Responsible team
- Frontline Lab editorial
- Input boundary
- Public sources and published material
- Automated work
- Collection, field checks, deduplication, summary and translation assistance
- Human responsibility
- Selection, verification, significance, and publication
Eight accountable stages from public source to publication
Each stage clarifies what automation does, what editors decide, and how readers can return to the evidence.
2. Frequency and field checks
Connectors run on task schedules and validate time, language, source name, and original URL. Public pages use immutable versions that have passed the publication workflow.
3. Deduplication and evidence links
Canonical URLs, content features, topics, and entities help identify duplicates. Cross-source items retain separate links rather than presenting several pieces of evidence as one source.
4. Summary and significance
Automation produces candidate summaries and analysis fields. Editors decide whether to use them, how to state them, and whether an item belongs in a priority layer. Frontline Lab summaries remain distinct from original sources.
5. Chinese and English translation
Chinese and English versions are stored and published separately. Review prioritizes facts, proper names, numbers, and tone. A page without a qualified counterpart does not enter the bilingual publication manifest.
6. Topic grouping and related content
Relationships prioritize shared events, entities, categories, tags, and community context. Topic and Item pages show only relationships that resolve to published entries and source evidence.
7. X and Reddit scope
X and Twitter coverage processes publicly accessible posts and account activity. Reddit coverage organizes public posts and available comments from high-signal communities. Neither promises platform-wide, unlimited-history, or every-comment coverage.
8. Human review and publication
Editors check titles, summaries, significance, source links, and bilingual pairs. Only reviewed versions enter the public manifest with a version, checksum, and updated time.
Editorial accountability
Corrections, removals, and responsibility
Published judgments must remain verifiable and correctable when the underlying evidence changes.
Method reviewed
How to request a correction
Record the page URL, the field that needs correction, and a verifiable original source, then send them to your existing Frontline Lab editorial contact. A public contact address is not yet configured on this information site; this page will not invent an email address or form in the meantime.
Source changes and removal
If a source is withdrawn, becomes unavailable, or is no longer suitable for publication, editors review the entry and revise or remove it through a later publication version. The source site's rights statements take priority.
Quotation and copyright
Public pages provide editorial summaries, necessary context, and descriptive source links. They do not replace the originating site by copying complete third-party articles. Readers should inspect the source for full context and terms.